Answers are grounded in this paper and cite their sources.
Clouds are usually present in optical satellite images, yet they occlude the ground truth, making it difficult to interpret satellite images. Moreover, distinguishing between areas with thick clouds and thin clouds is challenging when using the existing approaches. Therefore, in this article, an approach based on intensity inhomogeneity is proposed to differentiate thick clouds from thin clouds and vice versa. Intensity inhomogeneity indicates that the pixels in satellite images contain ground-truth information and other information, such as disturbances like clouds. In the proposed approach, intensity inhomogeneity is used to identify thin and thick clouds by assuming that intensity inhomogeneity is distributed as a Gaussian function in an image patch. Accordingly, the central pixels make more significant contributions than the other pixels. Intensity inhomogeneity can be approximated by incorporating it into an image segmentation model. Image segmentation models are numerically implemented in this article by using level-set functions in an iterative scheme. When convergence is reached, the given image is segmented, and simultaneously, thick cloud regions are distinguished from thin clouds based on the approximated intensity inhomogeneity values. Areas with approximated intensity inhomogeneity values smaller than one represent thin clouds. In this article, Sentinel-2 optical satellite images were used to evaluate the performance of the proposed approach. The processed results demonstrated that the proposed approach can be used to differentiate areas with thick clouds from those with thin clouds and vice versa. Based on this differentiation, various strategies can be applied to restore the ground truth of images covered with clouds.
No takes yet. Share an insight, caveat, or question.
Huang et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: